Skip to contentSkip to conversation

perf tests: Remove needless headers for bp_account

PATCH Applied Leo Yan 4 participants 12 messages Oct 18, 2019 – Oct 22, 2019 perf events

lore
Share & tools
lore

Leo Yan

author
#0 lore raw

PATCH 1/3 perf tests: Remove needless headers for bp_account

A few headers are not needed and were introduced by copying from other test file. This patch removes the needless headers for the breakpoint accounting testing.

---
 tools/perf/tests/bp_account.c | 4 ----
 1 file changed, 4 deletions(-)
tools/perf/tests/bp_account.c +0 -4
index 016bba2c142d..52ff7a462670 100644
@@ -10,11 +10,7 @@
 #include <unistd.h>
 #include <string.h>
 #include <sys/ioctl.h>
-#include <time.h>
 #include <fcntl.h>
-#include <signal.h>
-#include <sys/mman.h>
-#include <linux/compiler.h>
 #include <linux/hw_breakpoint.h>
 
 #include "tests.h"
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH 2/3 perf tests bp_account: Add dedicated checking helper is_supported()

Arm architecture supports the breakpoint accounting but it doesn't support breakpoint overflow signal handling. The current code uses the same checking helper, thus it disables both testings (bp_account and bp_signal) for arm platform.

For handling two testings separately, this patch adds a dedicated checking helper is_supported() for breakpoint accounting testing, thus it allows supporting breakpoint accounting testing on arm platform; the old helper test__bp_signal_is_supported() is only used to checking for breakpoint overflow signal testing.

---
 tools/perf/tests/bp_account.c   | 16 ++++++++++++++++
 tools/perf/tests/builtin-test.c |  2 +-
 tools/perf/tests/tests.h        |  1 +
 3 files changed, 18 insertions(+), 1 deletion(-)
tools/perf/tests/bp_account.c +16 -0
index 52ff7a462670..d0b935356274 100644
@@ -188,3 +188,19 @@ int test__bp_accounting(struct test *test __maybe_unused, int subtest __maybe_un
 
 	return bp_accounting(wp_cnt, share);
 }
+
+bool test__bp_account_is_supported(void)
+{
+	/*
+	 * PowerPC and S390 do not support creation of instruction
+	 * breakpoints using the perf_event interface.
+	 *
+	 * Just disable the test for these architectures until these
+	 * issues are resolved.
+	 */
+#if defined(__powerpc__) || defined(__s390x__)
+	return false;
+#else
+	return true;
+#endif
+}
tools/perf/tests/builtin-test.c +1 -1
index 55774baffc2a..8b286e9b7549 100644
@@ -121,7 +121,7 @@ static struct test generic_tests[] = {
 	{
 		.desc = "Breakpoint accounting",
 		.func = test__bp_accounting,
-		.is_supported = test__bp_signal_is_supported,
+		.is_supported = test__bp_account_is_supported,
 	},
 	{
 		.desc = "Watchpoint",
tools/perf/tests/tests.h +1 -0
index 72912eb473cb..9837b6e93023 100644
@@ -111,6 +111,7 @@ int test__map_groups__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
+bool test__bp_account_is_supported(void);
 bool test__wp_is_supported(void);
 
 #if defined(__arm__) || defined(__aarch64__)
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH v1 2/3] perf tests bp_account: Add dedicated checking helper is_supported()

Em Fri, Oct 18, 2019 at 04:55:30PM +0800, Leo Yan escreveu:

Quoted 10 lines

Arm architecture supports the breakpoint accounting but it doesn't support breakpoint overflow signal handling. The current code uses the same checking helper, thus it disables both testings (bp_account and bp_signal) for arm platform.

For handling two testings separately, this patch adds a dedicated checking helper is_supported() for breakpoint accounting testing, thus it allows supporting breakpoint accounting testing on arm platform; the old helper test__bp_signal_is_supported() is only used to checking for breakpoint overflow signal testing.

Looks sensible,

Applied, thanks.

- Arnaldo

Quoted 58 lines
---
 tools/perf/tests/bp_account.c   | 16 ++++++++++++++++
 tools/perf/tests/builtin-test.c |  2 +-
 tools/perf/tests/tests.h        |  1 +
 3 files changed, 18 insertions(+), 1 deletion(-)
tools/perf/tests/bp_account.c +16 -0
index 52ff7a462670..d0b935356274 100644
@@ -188,3 +188,19 @@ int test__bp_accounting(struct test *test __maybe_unused, int subtest __maybe_un
 
 	return bp_accounting(wp_cnt, share);
 }
+
+bool test__bp_account_is_supported(void)
+{
+	/*
+	 * PowerPC and S390 do not support creation of instruction
+	 * breakpoints using the perf_event interface.
+	 *
+	 * Just disable the test for these architectures until these
+	 * issues are resolved.
+	 */
+#if defined(__powerpc__) || defined(__s390x__)
+	return false;
+#else
+	return true;
+#endif
+}
tools/perf/tests/builtin-test.c +1 -1
index 55774baffc2a..8b286e9b7549 100644
@@ -121,7 +121,7 @@ static struct test generic_tests[] = {
 	{
 		.desc = "Breakpoint accounting",
 		.func = test__bp_accounting,
-		.is_supported = test__bp_signal_is_supported,
+		.is_supported = test__bp_account_is_supported,
 	},
 	{
 		.desc = "Watchpoint",
tools/perf/tests/tests.h +1 -0
index 72912eb473cb..9837b6e93023 100644
@@ -111,6 +111,7 @@ int test__map_groups__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
+bool test__bp_account_is_supported(void);
 bool test__wp_is_supported(void);
 
 #if defined(__arm__) || defined(__aarch64__)
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
merge bot Applied
#3 lore raw

[tip: perf/core] perf tests bp_account: Add dedicated checking helper is_supported()

In reply to Leo Yan
The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     e533eadf6596451880f518949cbb964dbd6189ae
Gitweb:        https://git.kernel.org/tip/e533eadf6596451880f518949cbb964dbd6189ae
Author:        Leo Yan <[email protected]>
AuthorDate:    Fri, 18 Oct 2019 16:55:30 +08:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Sat, 19 Oct 2019 15:35:01 -03:00

perf tests bp_account: Add dedicated checking helper is_supported()

The arm architecture supports breakpoint accounting but it doesn't support breakpoint overflow signal handling. The current code uses the same checking helper, thus it disables both testings (bp_account and bp_signal) for arm platform.

For handling two testings separately, this patch adds a dedicated checking helper is_supported() for breakpoint accounting testing, thus it allows supporting breakpoint accounting testing on arm platform; the old helper test__bp_signal_is_supported() is only used to checking for breakpoint overflow signal testing.

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Brajeswar Ghosh <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Michael Petlan <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Song Liu <[email protected]>
Cc Souptick Joarder <[email protected]>
Cc Will Deacon <[email protected]>
---
 tools/perf/tests/bp_account.c   | 16 ++++++++++++++++
 tools/perf/tests/builtin-test.c |  2 +-
 tools/perf/tests/tests.h        |  1 +
 3 files changed, 18 insertions(+), 1 deletion(-)
tools/perf/tests/bp_account.c +16 -0
index 52ff7a4..d0b9353 100644
@@ -188,3 +188,19 @@ int test__bp_accounting(struct test *test __maybe_unused, int subtest __maybe_un
 
 	return bp_accounting(wp_cnt, share);
 }
+
+bool test__bp_account_is_supported(void)
+{
+	/*
+	 * PowerPC and S390 do not support creation of instruction
+	 * breakpoints using the perf_event interface.
+	 *
+	 * Just disable the test for these architectures until these
+	 * issues are resolved.
+	 */
+#if defined(__powerpc__) || defined(__s390x__)
+	return false;
+#else
+	return true;
+#endif
+}
tools/perf/tests/builtin-test.c +1 -1
index 55774ba..8b286e9 100644
@@ -121,7 +121,7 @@ static struct test generic_tests[] = {
 	{
 		.desc = "Breakpoint accounting",
 		.func = test__bp_accounting,
-		.is_supported = test__bp_signal_is_supported,
+		.is_supported = test__bp_account_is_supported,
 	},
 	{
 		.desc = "Watchpoint",
tools/perf/tests/tests.h +1 -0
index 72912eb..9837b6e 100644
@@ -111,6 +111,7 @@ int test__map_groups__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
+bool test__bp_account_is_supported(void);
 bool test__wp_is_supported(void);
 
 #if defined(__arm__) || defined(__aarch64__)
author
#4 lore raw

PATCH 3/3 perf tests: Disable bp_signal testing for arm64

In reply to Leo Yan

As there have several discussions for enabling Perf breakpoint signal testing on arm64 platform; arm64 needs to rely on single-step to execute the breakpointed instruction and then reinstall the breakpoint exception handler. But if hook the breakpoint with a signal, the signal handler will do the stepping rather than the breakpointed instruction, this causes infinite loops as below:

         Kernel space              |            Userspace
-----------------------------------|--------------------------------
                                   |  __test_function() -> hit
				   |                       breakpoint
  breakpoint_handler()             |
    `-> user_enable_single_step()  |
  do_signal()                      |
                                   |  sig_handler() -> Step one
				   |                instruction and
				   |                trap to kernel
  single_step_handler()            |
    `-> reinstall_suspended_bps()  |
                                   |  __test_function() -> hit
				   |     breakpoint again and
				   |     repeat up flow infinitely

As Will Deacon mentioned [1]: "that we require the overflow handler to do the stepping on arm/arm64, which is relied upon by GDB/ptrace. The hw_breakpoint code is a complete disaster so my preference would be to rip out the perf part and just implement something directly in ptrace, but it's a pretty horrible job". Though Will commented this on arm architecture, but the comment also can apply on arm64 architecture.

For complete information, I searched online and found a few years back, Wang Nan sent one patch 'arm64: Store breakpoint single step state into pstate' [2]; the patch tried to resolve this issue by avoiding single stepping in signal handler and defer to enable the signal stepping when return to __test_function(). The fixing was not merged due to the concern for missing to handle different usage cases.

Based on the info, the most feasible way is to skip Perf breakpoint signal testing for arm64 and this could avoid the duplicate investigation efforts when people see the failure. This patch skips this case on arm64 platform, which is same with arm architecture.

[1] https://lkml.org/lkml/2018/11/15/205
[2] https://lkml.org/lkml/2015/12/23/477

---
 tools/perf/tests/bp_signal.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/tests/bp_signal.c +6 -9
index c1c2c13de254..166f411568a5 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.
+	 *
 	 * Just disable the test for these architectures until these
 	 * issues are resolved.
 	 */
-#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__)
+#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__) || \
+    defined(__aarch64__)
 	return false;
 #else
 	return true;
Signature · 1 line
2.17.1
#5 lore raw

Re: [PATCH v1 3/3] perf tests: Disable bp_signal testing for arm64

Em Fri, Oct 18, 2019 at 04:55:31PM +0800, Leo Yan escreveu:

Quoted 37 lines

As there have several discussions for enabling Perf breakpoint signal testing on arm64 platform; arm64 needs to rely on single-step to execute the breakpointed instruction and then reinstall the breakpoint exception handler. But if hook the breakpoint with a signal, the signal handler will do the stepping rather than the breakpointed instruction, this causes infinite loops as below:

         Kernel space              |            Userspace
-----------------------------------|--------------------------------
                                   |  __test_function() -> hit
				   |                       breakpoint
  breakpoint_handler()             |
    `-> user_enable_single_step()  |
  do_signal()                      |
                                   |  sig_handler() -> Step one
				   |                instruction and
				   |                trap to kernel
  single_step_handler()            |
    `-> reinstall_suspended_bps()  |
                                   |  __test_function() -> hit
				   |     breakpoint again and
				   |     repeat up flow infinitely

As Will Deacon mentioned [1]: "that we require the overflow handler to do the stepping on arm/arm64, which is relied upon by GDB/ptrace. The hw_breakpoint code is a complete disaster so my preference would be to rip out the perf part and just implement something directly in ptrace, but it's a pretty horrible job". Though Will commented this on arm architecture, but the comment also can apply on arm64 architecture.

For complete information, I searched online and found a few years back, Wang Nan sent one patch 'arm64: Store breakpoint single step state into pstate' [2]; the patch tried to resolve this issue by avoiding single stepping in signal handler and defer to enable the signal stepping when return to __test_function(). The fixing was not merged due to the concern for missing to handle different usage cases.

Quoted 4 lines

Based on the info, the most feasible way is to skip Perf breakpoint signal testing for arm64 and this could avoid the duplicate investigation efforts when people see the failure. This patch skips this case on arm64 platform, which is same with arm architecture.

Ok, applying,

- Arnaldo

Quoted 46 lines

[1] https://lkml.org/lkml/2018/11/15/205
[2] https://lkml.org/lkml/2015/12/23/477

---
 tools/perf/tests/bp_signal.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/tests/bp_signal.c +6 -9
index c1c2c13de254..166f411568a5 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.
+	 *
 	 * Just disable the test for these architectures until these
 	 * issues are resolved.
 	 */
-#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__)
+#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__) || \
+    defined(__aarch64__)
 	return false;
 #else
 	return true;
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
author
#6 lore raw

Re: [PATCH v1 3/3] perf tests: Disable bp_signal testing for arm64

Quoted 38 lines · Arnaldo Carvalho de Melo

Em Fri, Oct 18, 2019 at 04:55:31PM +0800, Leo Yan escreveu:

Quoted 37 lines

As there have several discussions for enabling Perf breakpoint signal testing on arm64 platform; arm64 needs to rely on single-step to execute the breakpointed instruction and then reinstall the breakpoint exception handler. But if hook the breakpoint with a signal, the signal handler will do the stepping rather than the breakpointed instruction, this causes infinite loops as below:

         Kernel space              |            Userspace
-----------------------------------|--------------------------------
                                   |  __test_function() -> hit
				   |                       breakpoint
  breakpoint_handler()             |
    `-> user_enable_single_step()  |
  do_signal()                      |
                                   |  sig_handler() -> Step one
				   |                instruction and
				   |                trap to kernel
  single_step_handler()            |
    `-> reinstall_suspended_bps()  |
                                   |  __test_function() -> hit
				   |     breakpoint again and
				   |     repeat up flow infinitely

As Will Deacon mentioned [1]: "that we require the overflow handler to do the stepping on arm/arm64, which is relied upon by GDB/ptrace. The hw_breakpoint code is a complete disaster so my preference would be to rip out the perf part and just implement something directly in ptrace, but it's a pretty horrible job". Though Will commented this on arm architecture, but the comment also can apply on arm64 architecture.

For complete information, I searched online and found a few years back, Wang Nan sent one patch 'arm64: Store breakpoint single step state into pstate' [2]; the patch tried to resolve this issue by avoiding single stepping in signal handler and defer to enable the signal stepping when return to __test_function(). The fixing was not merged due to the concern for missing to handle different usage cases.

Quoted 6 lines
Quoted 4 lines

Based on the info, the most feasible way is to skip Perf breakpoint signal testing for arm64 and this could avoid the duplicate investigation efforts when people see the failure. This patch skips this case on arm64 platform, which is same with arm architecture.

Ok, applying,

Thanks a lot, Arnaldo.

@Will, @Mark Rultland, very appreciate if you have time to review this patch and welcome any comments or suggestions! It's good you could confirm this patch so have more confidence for it.

Thanks,
Leo Yan

Quoted 52 lines

- Arnaldo

Quoted 46 lines

[1] https://lkml.org/lkml/2018/11/15/205
[2] https://lkml.org/lkml/2015/12/23/477

---
 tools/perf/tests/bp_signal.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/tests/bp_signal.c +6 -9
index c1c2c13de254..166f411568a5 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.
+	 *
 	 * Just disable the test for these architectures until these
 	 * issues are resolved.
 	 */
-#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__)
+#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__) || \
+    defined(__aarch64__)
 	return false;
 #else
 	return true;
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
merge bot Applied
#7 lore raw

[tip: perf/core] perf tests: Disable bp_signal testing for arm64

In reply to Leo Yan
The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     6a5f3d94cb69a185b921cb92c39888dc31009acb
Gitweb:        https://git.kernel.org/tip/6a5f3d94cb69a185b921cb92c39888dc31009acb
Author:        Leo Yan <[email protected]>
AuthorDate:    Fri, 18 Oct 2019 16:55:31 +08:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Sat, 19 Oct 2019 15:35:01 -03:00

perf tests: Disable bp_signal testing for arm64

As there are several discussions for enabling perf breakpoint signal testing on arm64 platform: arm64 needs to rely on single-step to execute the breakpointed instruction and then reinstall the breakpoint exception handler. But if we hook the breakpoint with a signal, the signal handler will do the stepping rather than the breakpointed instruction, this causes infinite loops as below:

         Kernel space              |            Userspace
  ---------------------------------|--------------------------------
                                   |  __test_function() -> hit
				   |                       breakpoint
  breakpoint_handler()             |
    `-> user_enable_single_step()  |
  do_signal()                      |
                                   |  sig_handler() -> Step one
				   |                instruction and
				   |                trap to kernel
  single_step_handler()            |
    `-> reinstall_suspended_bps()  |
                                   |  __test_function() -> hit
				   |     breakpoint again and
				   |     repeat up flow infinitely

As Will Deacon mentioned [1]: "that we require the overflow handler to do the stepping on arm/arm64, which is relied upon by GDB/ptrace. The hw_breakpoint code is a complete disaster so my preference would be to rip out the perf part and just implement something directly in ptrace, but it's a pretty horrible job". Though Will commented this on arm architecture, but the comment also can apply on arm64 architecture.

For complete information, I searched online and found a few years back, Wang Nan sent one patch 'arm64: Store breakpoint single step state into pstate' [2]; the patch tried to resolve this issue by avoiding single stepping in signal handler and defer to enable the signal stepping when return to __test_function(). The fixing was not merged due to the concern for missing to handle different usage cases.

Based on the info, the most feasible way is to skip Perf breakpoint signal testing for arm64 and this could avoid the duplicate investigation efforts when people see the failure. This patch skips this case on arm64 platform, which is same with arm architecture.

[1] https://lkml.org/lkml/2018/11/15/205
[2] https://lkml.org/lkml/2015/12/23/477

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Brajeswar Ghosh <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Michael Petlan <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Song Liu <[email protected]>
Cc Souptick Joarder <[email protected]>
Cc Will Deacon <[email protected]>
---
 tools/perf/tests/bp_signal.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/tests/bp_signal.c +6 -9
index c1c2c13..166f411 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.
+	 *
 	 * Just disable the test for these architectures until these
 	 * issues are resolved.
 	 */
-#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__)
+#if defined(__powerpc__) || defined(__s390x__) || defined(__arm__) || \
+    defined(__aarch64__)
 	return false;
 #else
 	return true;
#8 lore raw

Re: [tip: perf/core] perf tests: Disable bp_signal testing for arm64

Quoted 95 lines · On Mon, Oct 21, 2019 at 11:18:54PM -0000, tip-bot2 for Leo Yan wrote:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     6a5f3d94cb69a185b921cb92c39888dc31009acb
Gitweb:        https://git.kernel.org/tip/6a5f3d94cb69a185b921cb92c39888dc31009acb
Author:        Leo Yan <[email protected]>
AuthorDate:    Fri, 18 Oct 2019 16:55:31 +08:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Sat, 19 Oct 2019 15:35:01 -03:00

perf tests: Disable bp_signal testing for arm64

As there are several discussions for enabling perf breakpoint signal testing on arm64 platform: arm64 needs to rely on single-step to execute the breakpointed instruction and then reinstall the breakpoint exception handler. But if we hook the breakpoint with a signal, the signal handler will do the stepping rather than the breakpointed instruction, this causes infinite loops as below:

         Kernel space              |            Userspace
  ---------------------------------|--------------------------------
                                   |  __test_function() -> hit
				   |                       breakpoint
  breakpoint_handler()             |
    `-> user_enable_single_step()  |
  do_signal()                      |
                                   |  sig_handler() -> Step one
				   |                instruction and
				   |                trap to kernel
  single_step_handler()            |
    `-> reinstall_suspended_bps()  |
                                   |  __test_function() -> hit
				   |     breakpoint again and
				   |     repeat up flow infinitely

As Will Deacon mentioned [1]: "that we require the overflow handler to do the stepping on arm/arm64, which is relied upon by GDB/ptrace. The hw_breakpoint code is a complete disaster so my preference would be to rip out the perf part and just implement something directly in ptrace, but it's a pretty horrible job". Though Will commented this on arm architecture, but the comment also can apply on arm64 architecture.

For complete information, I searched online and found a few years back, Wang Nan sent one patch 'arm64: Store breakpoint single step state into pstate' [2]; the patch tried to resolve this issue by avoiding single stepping in signal handler and defer to enable the signal stepping when return to __test_function(). The fixing was not merged due to the concern for missing to handle different usage cases.

Based on the info, the most feasible way is to skip Perf breakpoint signal testing for arm64 and this could avoid the duplicate investigation efforts when people see the failure. This patch skips this case on arm64 platform, which is same with arm architecture.

[1] https://lkml.org/lkml/2018/11/15/205
[2] https://lkml.org/lkml/2015/12/23/477

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Brajeswar Ghosh <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Michael Petlan <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Song Liu <[email protected]>
Cc Souptick Joarder <[email protected]>
Cc Will Deacon <[email protected]>
---
 tools/perf/tests/bp_signal.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
tools/perf/tests/bp_signal.c +0 -8
index c1c2c13..166f411 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {

tools/perf/tests/bp_signal.c bool test__bp_signal_is_supported(void) · line 299

Quoted 6 lines
tools/perf/tests/bp_signal.c +3 -0
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.

Freudian slip?

Will

author
#9 lore raw

Re: [tip: perf/core] perf tests: Disable bp_signal testing for arm64

On Tue, Oct 22, 2019 at 02:14:25PM +0100, Will Deacon wrote:

[...]

Quoted 20 lines
Quoted 20 lines
tools/perf/tests/bp_signal.c +0 -8
index c1c2c13..166f411 100644
@@ -49,14 +49,6 @@ asm (
 	"__test_function:\n"
 	"incq (%rdi)\n"
 	"ret\n");
-#elif defined (__aarch64__)
-extern void __test_function(volatile long *ptr);
-asm (
-	".globl __test_function\n"
-	"__test_function:\n"
-	"str x30, [x0]\n"
-	"ret\n");
-
 #else
 static void __test_function(volatile long *ptr)
 {
@@ -302,10 +294,15 @@ bool test__bp_signal_is_supported(void)
Quoted 8 lines

tools/perf/tests/bp_signal.c bool test__bp_signal_is_supported(void) · line 299

Quoted 6 lines
tools/perf/tests/bp_signal.c +3 -0
 	 * stepping into the SIGIO handler and getting stuck on the
 	 * breakpointed instruction.
 	 *
+	 * Since arm64 has the same issue with arm for the single-step
+	 * handling, this case also gets suck on the breakpointed
+	 * instruction.

Freudian slip?

:D sorry for typo: s/suck/stuck.

Thanks for review and will send a patch to fix it.

Thanks,
Leo Yan

#10 lore raw
In reply to Leo Yan

Em Fri, Oct 18, 2019 at 04:55:29PM +0800, Leo Yan escreveu:

Quoted 3 lines

A few headers are not needed and were introduced by copying from other test file. This patch removes the needless headers for the breakpoint accounting testing.

Thanks, applied.

Quoted 23 lines
---
 tools/perf/tests/bp_account.c | 4 ----
 1 file changed, 4 deletions(-)
tools/perf/tests/bp_account.c +0 -4
index 016bba2c142d..52ff7a462670 100644
@@ -10,11 +10,7 @@
 #include <unistd.h>
 #include <string.h>
 #include <sys/ioctl.h>
-#include <time.h>
 #include <fcntl.h>
-#include <signal.h>
-#include <sys/mman.h>
-#include <linux/compiler.h>
 #include <linux/hw_breakpoint.h>
 
 #include "tests.h"
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
merge bot Applied
#11 lore raw
In reply to Leo Yan
The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     12d795637ba169650ea10ad6babcc5425255944a
Gitweb:        https://git.kernel.org/tip/12d795637ba169650ea10ad6babcc5425255944a
Author:        Leo Yan <[email protected]>
AuthorDate:    Fri, 18 Oct 2019 16:55:29 +08:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Sat, 19 Oct 2019 15:35:01 -03:00

perf tests: Remove needless headers for bp_account

A few headers are not needed and were introduced by copying from other test file. This patch removes the needless headers for the breakpoint accounting testing.

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Brajeswar Ghosh <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Michael Petlan <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Song Liu <[email protected]>
Cc Souptick Joarder <[email protected]>
Cc Will Deacon <[email protected]>
---
 tools/perf/tests/bp_account.c | 4 ----
 1 file changed, 4 deletions(-)
tools/perf/tests/bp_account.c +0 -4
index 016bba2..52ff7a4 100644
@@ -10,11 +10,7 @@
 #include <unistd.h>
 #include <string.h>
 #include <sys/ioctl.h>
-#include <time.h>
 #include <fcntl.h>
-#include <signal.h>
-#include <sys/mman.h>
-#include <linux/compiler.h>
 #include <linux/hw_breakpoint.h>
 
 #include "tests.h"